For reprint orders, please contact: REVIEW [email protected]

High-resolution DNA melting analysis for simple and efficient molecular diagnostics

Gudrun H Reed, High-resolution melting of DNA is a simple solution for genotyping, scanning Jana O Kent & and sequence matching. The melting profile of a PCR product depends on its GC content, Carl T Wittwer† length, sequence and heterozygosity and is best monitored with saturating dyes that †Author for correspondence Department of Pathology, fluoresce in the presence of double-stranded DNA. Genotyping of most variants is possible University of Utah Medical by the melting temperature of the PCR products, while all variants can be genotyped with Center, 5B418, unlabeled probes. Mutation scanning and sequence matching depend on sequence 50 North Medical Drive, Salt Lake City, UT 84132, differences that result in heteroduplexes that change the shape of the melting curve. USA High-resolution DNA melting has several advantages over other genotyping and scanning Tel.: +1 801 581 4737; methods, including an inexpensive closed tube format that is homogenous, accurate and Fax: +1 801 581 6001; rapid. Owing to its simplicity and speed, the method is a good fit for personalized medicine E-mail: carl.wittwer@ path.utah.edu as a rapid, inexpensive method to predict therapeutic response.

High-resolution melting is a new method for Thermal melting of DNA was historically DNA analysis introduced in 2002 by a collabo- monitored by UV absorbance. For high-quality ration between academics (University of Utah, melting curves, µg amounts of DNA and rates of UT, USA) and industry (Idaho Technology, UT, 0.1–1.0°C/min were required. In contrast to USA). As the simplest method for genotyping, absorbance, fluorescence analysis of DNA melt- mutation scanning and sequence matching, its ing is more sensitive, and only nanogram popularity is growing. No separations or process- amounts are needed, conveniently provided by ing of the samples is required. After PCR ampli- PCR amplification. Methods that monitor fication, melting curves are generated by DNA melting by fluorescence have become pop- monitoring the fluorescence of a saturating dye ular with the advent of real-time PCR [4] and that does not inhibit PCR. were introduced 10 years ago with the Light- When combined with rapid-cycle PCR [1], Cycler® [5–7]. Capillary sample formats and high-resolution melting is an ideal solution for smaller sample volumes allowed better tempera- personalized DNA diagnostics. For example, ture control, enabling much faster melting rates warfarin is a commonly used anticoagulant with of 0.1–1.0°C/s. SYBR® Green I was introduced a narrow therapeutic range. If the dose is not as a sensitive, convenient dye for PCR product right, either serious bleeding or clotting may melting analysis. occur. The required dose of warfarin is modified by sequence variants in genes that affect its High-resolution DNA melting with metabolism. The genotyping of three loci saturation dyes explains much of the variance in the required Modern high-resolution DNA melting is dose [2]. Rapid genotyping to determine appro- enabled by novel saturation dyes and high- priate dosing can be critical in emergency surgery. resolution instruments. With SYBR Green I, it is Rapid-cycle PCR (<15 min) followed by high- difficult to guarantee saturation of the PCR resolution melting (<2 min) provides a rapid product with dye as only limited concentrations solution. To give another example, a patient with can be used before it inhibits PCR. Although typhoid fever requires rapid treatment. However, single-base genotyping with SYBR Green I has there are genetic variants of Salmonella that been reported [8–10], the results have been Keywords: DNA melting, result in resistance to the commonly used anti- questioned [11] and, in our hands, is not genotyping, heteroduplexes, biotics. These variants can be detected by high- robust [12,13]. Much better results are possible HLA matching, LCGreen® dye, melting temperature, resolution melting in order to direct alternative with a new generation of saturation dyes, specifi- mutation scanning antibiotic therapy. Not only are these methods cally developed for high-resolution melting. fast, but they are inexpensive because real-time These dyes, under the tradename LCGreen®, part of thermal cyclers and covalently-labeled probes are are compatible with PCR over a wide range of not required [3]. concentrations. Single-base variants and small

10.2217/14622416.8.6.597 © 2007 Future Medicine Ltd ISSN 1462-2416 Pharmacogenomics (2007) 8(6), 597–608 597 REVIEW – Reed, Kent & Wittwer

insertions or deletions are easily detected and instruments dedicated to high-resolution melt- genotyped with LCGreen dyes. Alternative satu- ing perform better than real-time instruments rating dyes other than the LCGreen family adapted to high-resolution melting [24–26]. (LCGreen I and LCGreen Plus) are beginning to However, there is convenience in having both appear [14,15], although no comparative studies functions (amplification and melting analysis) are available. combined in one instrument, and some prefer to interpret melting data in the context of real- High-resolution DNA time data. Nevertheless, integrated real-time melting instruments data comes at a cost. The HR-1 remains the In addition to saturation dyes, new instrumenta- gold standard in melting quality and the Light- tion was necessary to fully empower high-resolu- Scanner provides the highest throughput. Melt- tion melting techniques. The first high- ing resolution is directly correlated to resolution melting instrument (HR-1, Idaho performance, that is, scanning sensitivity and Technology) was developed with the goal of specificity and genotyping accuracy. Most con- making DNA melting as precise and accurate as ventional real-time thermal cyclers do not per- possible in order to investigate the potential of form well compared with high-resolution the technique. Single samples are analyzed in instruments. The detailed technical perform- LightCycler capillaries surrounded by a metal ance of 16 different melting instruments was ingot heated by a resistance coil. Amplification is recently compared in a series of reports [24–26]. performed in a LightCycler (Roche) or the High-resolution melting methods have been low-cost RapidCycler II (Idaho Technology). compared with other techniques in recent Analysis is rapid (1–2 min) for a throughput of reviews [27–30]. In what follows, the fundamen- approximately 45 samples/h. At 0.3°C/s, approx- tals of DNA melting analysis will first be cov- imately 65 points are acquired and plotted per °C ered, followed by applications, including without any smoothing of the data. Geno- testing for known sequence variants (geno- typing [12,13], mutation scanning [16–18] and typing), identifying similarities or dissimilari- sequence matching [19] were all first demon- ties in DNA (sequence matching) and strated on this instrument. Demand for a 96- or screening for (scanning). 384-well plate format led to the introduction of the LightScanner® (Idaho Technology) [20–23]. Fundamentals of fluorescent DNA Paired with standard plate thermal cyclers, the melting analysis throughput of such a system is very high with Certain dyes fluoresce strongly in the presence many thermal cyclers funneling into one Light- of double-stranded DNA. The most familiar of Scanner. With a melting turnaround of 5 min, these is ethidium bromide, giving the red bands over 4000 samples can be analyzed per hour on a often observed in electrophoresis gels. Asym- 384-well LightScanner if enough thermal cyclers metric cyanine dyes, such as SYBR Green I and are available. LCGreen, are even brighter and are the dyes of Recently, some real-time thermal cyclers have choice in fluorescence melting analysis and real- been modified to incorporate high-resolution time PCR [4]. In order to generate a melting melting, including the LC480 (Roche) and the curve, the sample is heated through a range of Rotor-Gene 6000 (Corbett). These instruments temperatures, while fluorescence is continu- approach high-resolution data quality by melt- ously collected (Figure 1A). Any double-stranded ing at slower rates. For example, the HR-1 DNA present will fluoresce strongly at low tem- melts at 0.3°C/s, taking just over a minute to peratures. As the temperature is increased, the pass through a 20°C range. In contrast, the fluorescence will decrease, at first slowly, and Rotor-Gene temperature ramps are defined as then, at a characteristic temperature the fluores- ‘°C’. Under recommended conditions (0.1°C), cence rapidly drops, reflecting the melting of a 2s hold is performed at each 0.1°C step in DNA into single strands. The melting tempera- temperature. The resulting actual measured rate ture (Tm) of a DNA duplex is characteristic of is 0.017°C/s, producing 10 points/°C and its GC content, length and sequence and is the requiring 20 min for the same temperature temperature at which the normalized fluores- range. Although this is 18-times slower than cence is 50% (Figure 1B). Accurate calculation of the HR-1, the extra time is necessary to Tm first requires background removal before improve the data quality. Not surprisingly, normalization. The major component of back- comparative studies indicate that, in general, ground is linear and arises from a physical

598 Pharmacogenomics (2007) 8(6) futurefuture sciencescience ggrouproup High-resolution DNA melting analysis for molecular diagnostics – REVIEW

Cost Figure 1. Fluorescent DNA melting analysis. Cost advantages of high-resolution melting are derived from the simplicity of the technique. A Original data The only reagent required is a saturating DNA 600 dye that costs less than the PCR reagents/con- tainer. Available hardware ranges from real-time instruments at more than US$50,000, down to 400 $15,000 [3]. All instrument options cost signifi- Double-stranded cantly less than a denaturing high-pressure liquid Single-stranded DNA DNA (random coils) chromatography (DHPLC) setup for mutation scanning. Additional cost advantages include the

Fluorescence 200 time saved and errors avoided because the method is closed-tube, and the ability to perform both genotyping and scanning on one platform 0 with one generic reagent. 76 79 82Tm 85 88 91 Temperature (°C) Workflow The saturating dye is added into the PCR before B Normalized melting curve amplification, so no sample processing or addi- 100 tions are necessary after PCR has begun. DNA extraction and quantification are usually per- formed before PCR. For best results, all test and 75 control DNA should be prepared in the same way Double-stranded and added into the PCR at the same concentra- DNA Single-stranded tion. However, good results can also be obtained 50 DNA (random coils) from crude DNA preparations, such as those pre- pared from dried blood spots without quantifica- 25 tion [32]. If care is taken to prevent undesired side reactions through PCR optimization and the Fluorescence (normalized) reaction is run into the plateau phase for all sam- 0 Tm 76 79 8285 88 91 ples, the initial DNA concentration can vary Temperature (°C) between samples by at least 100-fold.

(A) Original fluorescence data showing a linear decrease of fluorescence at low PCR optimization temperature, followed by a rapid decrease centered around the melting temperature (Tm). Fluorescence is low when the DNA is single stranded. (B) The Robust, specific PCR is critical when results original data is normalized between 0 and 100% after background subtraction depend on the PCR product melting profile. Use so that the curve is horizontal outside of the transition. of a gradient thermal cycler and gel electro- phoresis is still one of the best methods for opti- attribute of fluorescence: as the temperature is mization of conditions, and varying the Mg2+ raised, fluorescence decreases. At lower temper- concentration usually allows multiple targets to atures, an exponential component of back- be amplified under identical conditions. ground becomes apparent that arises from dye binding to high concentrations of primers. Genotyping Methods to remove linear [4] and exponential [31] Although there are many methods of genotyping, background have been described and are closed-tube methods have strong advantages for incorporated into commercial high-resolution the clinical laboratory, point-of-care diagnostics melting software. and personalized medicine. Since no processing is The Tm of a PCR product is a convenient required between amplification and analysis, the metric, but it is only one point on the melting need for automation and risk of contamination curve. More information is contained in the are eliminated. These methods conventionally use complete melting curve than in the Tm. The -specific labeled probes, often a fluorescent shape of the melting curve is used extensively in dye and a quencher that separate during amplifi- sequence matching and mutation scanning as cation by hydrolysis and/or loss of secondary an indicator of heteroduplexes formed from structure [28]. In order to genotype correctly, two heterozygous DNA. probes, one matching the wild-type sequence and

futurefuture sciencescience groupgroup www.futuremedicine.com 599 REVIEW – Reed, Kent & Wittwer

another matching the mutation sequence, are usu- as the size and the distance from the ally required. Typically, the fluorescence is meas- labeled primer increased. For detection, the ured in a real-time PCR machine once each cycle labeled primer had to be in the same melting during annealing or extension. domain as the sequence variant. This problem was solved in 2003 with the introduction of Genotyping by melting saturation dyes [13]. Genotyping by closed-tube melting analysis was With saturation dyes, the PCR product is introduced in 1997 [33]. The method is inher- labeled along its entire length, so that all melting ently more powerful than allele-specific methods domains are detected. This is demonstrated in as many different are distinguished and Figure 2B, where all genotypes of a C/T single hybridization is monitored over a range of tem- base variant in a two-domain melting curve are peratures, rather than only at a single tempera- shown. The difference between genotypes is ture. Before the advent of high-resolution revealed in the lower temperature domain, while melting analysis, labeled probes were usually neces- the upper melting domain is constant between sary for single-base genotyping by melting. Either genotypes. The differences between genotypes fluorescence resonance energy transfer [33,34] or are greater for smaller (Figure 2A) than guanosine quenching [35] produced the probe for large amplicons (Figure 2B). melting curves necessary for genotyping. For single-base genotyping, heterozygotes are Depending on the sequence under the probe, easy to identify because of the change in curve different alleles resulted in different probe melt- shape. However, not all homozygotes can be dis- ing temperatures. Heterozygous PCR products tinguished by Tm [38]. Approximately 84% of all were easily distinguished from homozygous sam- human single-base changes result in an A:T to ples by a double peak on derivative melting curve G:C interchange with a Tm difference of approx- plots. Both fluorescence color and Tm were imately 1°C in small amplicons. In the remain- exploited for multiplexing [36]; for example, ing 16%, the is inverted or neutral (A:T genotyping of HbC, HbS and HbE of human to T:A or G:C to C:G) and the Tm difference is β-globin [37]. smaller. In approximately 4% of human single base changes, nearest-neighbor symmetry pre- Genotyping of PCR products by dicts no difference in Tm. In such a case, mixing high-resolution amplicon melting is necessary for complete genotyping. If mixing High-resolution melting analysis enables geno- is performed after PCR is complete, a known typing without probes, even when the sequence homozygote is mixed into each unknown change is only a single base. Consider an A>C homozygote and the mixture melted again. variation with possible genotypes A/A, A/C and Alternatively, a known genotype can be added C/C (Figure 2A). If a small amplicon is generated into all samples before PCR and quantitative with PCR primers that bracket the variable heteroduplex analysis is performed [39]. locus, all three genotypes are easily distin- Different heterozygotes can often be distin- guished. The A/A and C/C curves are similar in guished from each other by differences in curve shape with the Tm of the C/C homozygote shape. In one study, all 21 random pairs of approximately 1°C higher than that of the A/A unique heterozygotes were distinguishable by homozygote. The melting curve of the A/C high-resolution melting of small amplicons [40]. heterozygote differs in shape from that of the In another study of 24 exons in two genes, all homozygotes with a more gradual transition common variants were distinguishable from over a larger temperature range. The greater disease-causing variants and each other [41]. range results from melting four different However, not all heterozygotes can be distin- duplexes: two homoduplexes (A/A and C/C) guished. For example, identical nearest-neigh- and two heteroduplexes (A/C and C/A). bor changes may occur at different locations High-resolution genotyping without probes within the same amplicon, such as the same (direct PCR product genotyping) was first mutation at different cysteine residues in the reported using fluorescently-labeled primers [12]. RET proto-oncogene [42]. A 113 bp fragment of β-globin was amplified High-resolution amplicon melting has been bracketing the HbS, HbC, and HbE single-base applied to both human (diploid) and microbial loci. All homozygotes (AA, SS, CC and EE) and (monoploid) genotyping. Human targets include heterozygotes (AS, AC, AE and SC) tested were disease-associated variants in β-globin [12,13,38], distinguished. Genotyping became more difficult cystic fibrosis [12,13,23,43], factor V [23,38],

600 Pharmacogenomics (2007) 8(6) futurefuture sciencescience ggrouproup High-resolution DNA melting analysis for molecular diagnostics – REVIEW

Figure 2. Genotyping by amplicon melting. duplexes will be saturated with dye, giving melting regions for both the probe and the amplicon. Such a composite melting curve for A Small amplicon melting factor V Leiden genotyping is shown in 100 Figure 3A. Each region alone provides unambigu- ous genotyping. Considered together, cross-val- 80 A/C C/C idation provides an additional level of confidence. When only the unlabeled probe 60 region is considered, the melting curves are usually plotted as derivative plots (Figure 3B). 40 Genotyping with unlabeled probes was first published in 2004 [50]. To prevent polymerase 20 A/A extension, the probes are usually blocked at the

Fluorescence (normalized) 3´-end, often with phosphate, although other 0 blockers are more stable [51]. High-resolution 78 80 8284 86 88 melting improves the quality of the melting Temperature (°C) curves and allows more variants to be B Two-domain genotyping distinguished from each other. However, 100 unlabeled probe genotyping can be performed on lower resolution instruments, including the 80 LightTyper® and the LightCycler [50], as long as appropriate data analysis software is available [31]. 60 TT The probes are usually present during PCR, although they can be added after amplification is 40 complete without breaking the closed-tube C/C environment [52]. 20 C/T Unlabeled probe analysis allows fine dis- T/T Fluorescence (normalized) crimination of variants under the probe. 0 Probes can be designed to mask certain variants 82 83 84 85 Temperature (°C) or segments by incorporating deletions, mis- matches or universal bases [53]. Multiple (A) Single-base genotyping by small amplicon melting. (B). Single-base unlabeled probes can interrogate different genotyping within a large DNA fragment (544 bp) that melts in two domains. Panel B is reprinted with permission of AACC [13]. amplified regions. For example, two unlabeled probes strategically positioned within exon 10 prothrombin [38], 5,10-methylenetetrahydro- of the cystic fibrosis gene were used to geno- folate reductase [38,44] and hemochromatosis type six different variants [23]. Unlabeled proteins [38,39], platelet antigens [21,45], lactase [21], probes are helpful when amplicon melting cytochrome P450 2C9 [46] and methylation of alone does not provide adequate detail in the MGMT promoter region [15]. Microbial highly polymorphic regions. Unlabeled probe targets include mycobacterial typing using genotyping can also be combined with scan- hsp65 [47], bacterial speciation using the 16s ning for unknown variants [23]. Any sequence rRNA gene [48], identifying gyrA mutations that variation between the primers will affect cause quinolone resistance in Salmonella [3] and amplicon melting, while only a variant under Aspergillus speciation [49]. the probe will affect probe melting. Unlabeled probe and amplicon genotyping Unlabeled probe genotyping were recently compared [21]. Unlabeled probe An interesting variation on genotyping with genotyping was successful in all cases. By contrast, saturation dyes is to include an unlabeled successful amplicon genotyping depends on the probe. In addition to the full-length PCR prod- Tm difference between homozygotes [38,39] and uct, the probe produces additional melting data the instrument resolution [24–26]. Genotyping focused on the region under the probe. accuracy is better with smaller amplicons as the Unequal primer concentrations are used to gen- Tm differences are magnified and the chance of erate one strand of DNA in excess. Some of the unexpected variants between the primers is less. excess strand hybridizes to the complementary Most variants can be directly genotyped by ampli- unlabeled probe. Both probe and amplicon con melting, but a small minority requires mixing

futurefuture sciencescience groupgroup www.futuremedicine.com 601 REVIEW – Reed, Kent & Wittwer

Figure 3. Simultaneous genotyping and scanning. repeat typing that would be rapid and amenable to high-throughput analysis, while minimizing the danger of contamination. High-resolution melting has been used to A Composite melting curves detect 6–102 bp internal tandem duplications in 100 Probe the juxtamembrane domain of the FMS-like tyro- region Amplicon 80 region sine kinase 3 gene that are associated with acute myelogenous leukemia [60]. Internal duplications 60 were identified based on variation in the melting curve shape compared with wild-type amplicons. 40 More difficult than the detection of the pres- Wild-type …AGGCGAGGA… ence of duplications is complete genotyping of 20 R506QHet …– – – –A– – – –… R506QHom …– – – –A– – – –… short tandem repeats. For example, can high-reso-

Fluorescence (normalized) 0 lution melting be used to eliminate the need for 58 62 66 70 74 78 82 electrophoresis in repeat genotyping? Although Temperature (°C) this remains a difficult problem, Figure 4 shows progress toward a solution [Unpublished Data]. A poly- morphic tetranucleotide repeat region was B Unlabeled probe genotyping amplified with primers just outside the repeats. 24 Resulting normalized and derivative melting curves are shown in Figure 4. The difference in sta- 18 bility between hetero- and homoduplexes appears 12 proportional to the size difference between the two alleles. These results show that at least some

-dF/dT 6 genotypes can be differentiated from each other, although complete genotyping of highly 0 polymorphic repeats remains a future goal. -6 58 60 62 64 66 68 70 Sequence matching Temperature (°C) In some cases, complete genotyping of the target (A) Composite normalized melting curve showing transitions in both the probe DNA is less important than determining whether and amplicon regions (factor V Leiden locus). (B) Derivative plot of the probe DNA sequences match. This scenario occurs in region from part A after background removal by exponential subtraction. tissue transplantation, genotype–phenotype cor- Portions of this figure are reprinted with permission of AACC [23]. relation and forensics. That is, sequence know- ledge of the genotype is not needed, but sequence with a known genotype [39]. Internal temperature identity is. For example, in living-related organ controls can be used to improve amplicon geno- transplantation, siblings are usually genotyped typing accuracy, especially on lower resolution for HLA to obtain the best major histo- instruments [21,44]. compatibility match. This involves serotyping or Targets genotyped with unlabeled probes genotyping at several loci, usually HLA A, B, C include factor V [23,54], cystic fibrosis [23,50], and DR by often laborious means. However, human platelet antigens [21], the RET proto- what is really important is to find a compatible oncogene [42], lactase [21] and hereditary sibling, that is, one with HLA sequence identity. hemorrhagic telangiectasia [41]. For each available sibling, there is a 25% chance of a complete match. Repeat typing HLA sequence identity (matching) by high- Tandem repeats are scattered throughout both resolution melting was demonstrated using the eukaryotic and prokaryotic genomes and are highly polymorphic HLA-A locus in all seven highly polymorphic. A number of techniques cases of shared alleles among two individuals [19]. have been used for repeat typing, such as gel and HLA genotype identity was suggested when two capillary electrophoresis [55], capillary arrays [56], individuals had the same melting curves. Iden- microchip capillary electrophoresis [57], mass tity was confirmed by comparing the melting spectrometry [58] and hybridization arrays [59]. curve of a 1:1 mixture with the individual melt- High-resolution melting of PCR-amplified ing curves. If the samples are not identical, dif- repetitive regions is an interesting option for ferent heteroduplexes are formed that change the

602 Pharmacogenomics (2007) 8(6) futurefuture sciencescience ggrouproup High-resolution DNA melting analysis for molecular diagnostics – REVIEW

Figure 4. Genotyping a tetranucleotide repeat. Some of the methods are manual and labor inten- sive, while others are complex and require special- ized instrumentation. Many are based on detection A Normalized melting curves of heteroduplexes (mismatched duplexes) formed after amplification of heterozygous DNA. The 100 [13,13] [12,13] need for processing the sample after PCR is a [11,13] severe disadvantage. 75 [10,13] High-resolution melting analysis is a scanning method that does not require any processing, 50 reagent additions or separations after PCR. Ideal C melting rates are 0.1–0.3°C/s, so that the analy- 25 [13,13] sis is usually complete in 1–5 min. The sensitiv- ity and specificity are better than DHPLC [43]. 0 Fluorescence (normalized) Single-base changes, insertions and deletions can [12,13] all be detected, as long as the PCR primers B Derivative melting curves bracket the variation. This limitation is similar to sequencing: deletions of entire genes and 40 exons will usually go undetected. [11,13] Mutation scanning by high-resolution melting 30 depends on the melting of heteroduplexes that distort the shape of the melting curve (Figure 6). 20 This distortion can be seen by comparing the dF/dT - - normalized melting curves of a homozygous 10 [10,13] standard to a heterozygous sample [12,13]. In order to focus on comparing curve shape, the bottoms 0 of the curves are superimposed by shifting the 68 71 74 77 curves along the temperature axis until they are Temperature (°C) overlaid (Figure 6A). As the difference between (A) Normalized, temperature-shifted melting curves of locus D5S818 with the curves is small, it is often magnified by plotting number of repeats indicated. (B) Derivative plots of the same samples. the difference between samples (Figure 6B). Each (C) Duplexes formed by the four genotypes tested. curve is usually subtracted point-by-point from the homozygous reference (or an average of all shape of the melting curve. The potential to wild-type curves analyzed). Although difference reduce a very complex genotyping problem to a curves look similar to derivative melting curves simple, closed-tube, rapid process is attractive. (Figure 6C), they should not be confused. Deriva- tive curves are commonly used in melting curve Scanning for sequence variants by genotyping [33,34]. However, because they require high-resolution melting analysis data smoothing, derivative curves in high-resolu- Many methods for mutation scanning (as opposed tion display should be used cautiously, despite to genotyping) have been developed to screen for their familiarity. differences between the two copies of DNA within The sensitivity and specificity of scanning for an individual (Figure 5). These techniques include heterozygous single-base changes were system- single-strand conformational anal- atically studied using a set of engineered plas- ysis (SSCP) [61], denaturing gradient gel electro- mids [17]. All possible base changes were phoresis (DGGE) [62], DHPLC [63], temperature considered in PCR products from 50 bp to 1 kb gradient capillary electrophoresis (TGCE) [64] and in a background of 40, 50 or 60% GC content. even mass spectroscopy [65]. Sequencing provides For PCR products less than 400 bp, sensitivity both genotyping and scanning at the same time, and specificity were 100%. In PCR products but requires extensive automation, instrument- 400–1000 bps in length, sensitivity was 96.1% ation and analysis. All of these methods require with a specificity of 99.4%. The position of the separation of the sample on a gel or other matrix, variant within the PCR product did not affect some after additional processing, enzymatic or scanning accuracy. chemical reactions. Any processing increases the Although designed to detect heterozygotes, risk of contamination in future reactions because high-resolution scanning often detects homo- PCR products are exposed to the environment. zygyous changes as well. As discussed previously,

futurefuture sciencescience groupgroup www.futuremedicine.com 603 REVIEW – Reed, Kent & Wittwer

Figure 5. Workflow of different mutation scanning methods.

DGGE: Denaturing gradient gel electrophoresis; DHPLC: Denaturing high-pressure liquid chromatography; SSCP: Single-strand conformational polymorphism analysis; TGCE: Temperature gradient capillary electrophoresis.

96% of human single-base changes have homo- curves of the same genotype can be mathemati- zygotes that differ in Tm and should be detectable. cally clustered together, eliminating the guesswork What is more surprising is that most homozygotes of genotype assignment. In the large majority of are detectable by curve shape changes alone, that cases, common polymorphisms can be eliminated is, after temperature shifting has been performed. by amplicon melting alone. Secondary genotyping Even though many homozygotes can be detected, or sequencing is seldom necessary. it is still wise to mix an unknown sample with a Mutation scanning by high-resolution melting known wild-type sample for detection of has been reported for c-kit [18,66–68], medium- hemizygous variants (X-linked or Y chromosome) chain acyl-CoA dehydrogenase [16], primary car- or if homozygous variants are likely. nitine deficiency [69], RET [42,70], epidermal The need for controls is controversial. The cau- growth factor receptor [71–74], exostoses 1 and 2 [22], tious will include wild-type controls, but they are gap junction protein β1 [20], K-ras [14,74], phenyl- not necessary when variants are rare and many alanine hydroxylase [32], v-raf murine sarcoma samples are analyzed. The suspicious will include viral oncogene homolog B1 [66,74], p53 [74], negative controls without template, although such HER2 [71,72], hereditary hemorrhagic telangi- controls must be checked at the original data stage ectasia [41], and some exons of the cystic fibrosis and not normalized (you cannot normalize a gene [23,43]. melting transition when it is not there). The com- pulsive will include positive controls, even though Future perspective the variants identified will most likely be different High-resolution DNA melting provides very sim- from the positive controls included. One place ple solutions for genotyping, sequence matching where positive controls are useful is in the identifi- and mutation scanning. The technique is new, cation of common polymorphisms, that is, vari- but expanding rapidly as high-resolution instru- ants that are not of interest. As discussed above ments and dyes become available. As fluores- under sequence matching, identical amplicon cently-labeled probes and real-time PCR are not melting curves are strong evidence of sequence required, high-resolution methods have cost and identity. Identity can be confirmed by mixing simplicity advantages over other closed-tube with a standard and re-melting, small amplicon genotyping approaches. Those few variants that genotyping, unlabeled probe genotyping, or cannot be identified by amplicon melting can be sequencing. In one example of scanning genotyped with unlabeled probes. The complex- 24 exons [41], benign polymorphisms were present ity and cost of labeled probes for genotyping is in 96% of normal samples, greatly reducing the destined to make them obsolete. positive predictive value of mutation detection. Sequence matching by high-resolution melting When common polymorphisms were identified by can be used when exact genotyping is not neces- amplicon melting, the positive predictive value for sary. For example, siblings considering living- mutation detection increased to 100%. Melting related organ transplantation can be rapidly and

604 Pharmacogenomics (2007) 8(6) futurefuture sciencescience ggrouproup High-resolution DNA melting analysis for molecular diagnostics – REVIEW

Figure 6. Melting curve plots useful in mutation scanning. inexpensively matched for HLA compatibility, replacing laborious, expensive genotyping meth- ods. Specific genotypes correlate to melting curve A Normalized, temperature-shifted melting curves shape and position, so that genetic variation can 100 WT A be visualized on a two-dimensional difference plot T HET for genotype–phenotype correlation. Finally, 75 A T identity may be established through melting anal- G ysis of variable regions such as HLA or single-base 50 C changes, although requirements of DNA purity A and quality may limit practical applications. 25 C High-resolution melting is currently the best G method for mutation scanning because no

Fluorescence (normalized) processing or separations are required and the 0 T cost is minimal. When implemented correctly, B Difference curves 95–99% of the need for sequencing disappears. 20 Since all sequencing first requires PCR amplifi- cation, high-resolution melting can be inserted 15 into the sequencing process. Mutation scanning by melting is nondestructive, so that any positive 10 samples can be further processed for sequencing if simpler methods for identification (matching 5 and genotyping) fail. So, what does the future hold? High-resolution

Fluorescence difference melting can be extended to interrogate RNA 0 sequence variability by standard reverse-tran- -6 scriptase PCR. In combination with real-time C Derivative curves PCR, detection, quantification and genotyping are 45 all feasible on the same sample in one assay. Such analysis could be applied, for example, to hepatitis C where detection, quantification and genotyping 30 are all clinically relevant. Application to in situ PCR has not been explored but may be feasible. The possibility of direct detection without PCR is -dF/dT 15 also unexplored, but may be possible for high copy number plasmids or double-stranded viruses.

0 Disclosure 78 80 82 84 86 Aspects of high-resolution melting are licensed from the Uni- Temperature (°C) versity of Utah to Idaho Technology. CTW has equity interest HET: Heterozygous DNA (A>G single-base change); WT: Wild-type homozygote. in Idaho Technology. GHR and JOK have nothing to disclose.

Executive summary

• High-resolution DNA melting provides the simplest methods for genotyping and mutation scanning.

• High-resolution DNA melting provides rapid analysis (1–5 min after PCR) without reagent additions or separations.

• The resolution of melting instrumentation is critical for accuracy, sensitivity and specificity – most real-time thermal cyclers do not perform well.

• The DNA dye used is critical – heteroduplex detection is enabled with saturating double-stranded DNA dyes.

• Fluorescently-labeled probes and real-time PCR are not required.

• Genotyping methods include amplicon melting and unlabeled probes.

• High-resolution melting analysis is a closed-tube approach for fast, accurate and high-throughput mutation scanning.

• High-resolution melting analysis offers low-cost analysis compared with many alternatives.

futurefuture sciencescience groupgroup www.futuremedicine.com 605 REVIEW – Reed, Kent & Wittwer

Bibliography 9. Lipsky RH, Mazzanti CM, Rudolph JG 20. Kennerson ML, Warburton T, Nelis E et al.: Papers of special note have been highlighted as et al.: DNA melting analysis for detection of Mutation scanning the GJB1 gene with either of interest (•) or of considerable interest (••) single nucleotide polymorphisms. Clin. high-resolution melting analysis: to readers. Chem. 47, 635–644 (2001). implications for mutation scanning of genes 1. Wittwer CT, Herrmann MG: Rapid thermal 10. Dufresne SD, Belloni DR, Wells WA, for Charcot-Marie-Tooth disease. Clin. cycling and PCR kinetics. In: PCR Methods Tsongalis GJ: BRCA1 and BRCA2 mutation Chem. 53, 349–352 (2007). Manual. Innis M, Gelfand D, Sninsky J screening using SmartCycler II high- 21. Liew M, Seipp M, Durtschi J et al.: (Eds). Academic Press, San Diego, CA, USA resolution melt curve analysis. Arch. Pathol. Closed-tube SNP genotyping without 211–229 (1999). Lab. Med. 130, 185–187 (2006). labeled probes/a comparison between • Review of rapid temperature cycling 11. von Ahsen N, Oellerich M, Schutz E: unlabeled probe and amplicon melting. Am. for PCR. Limitations of genotyping based on J. Clin. Pathol. 127, 1–8 (2007). 2. Carlquist JF, Horne BD, Muhlestein JB amplicon melting temperature. Clin. Chem. 22. Lonie L, Porter DE, Fraser M et al.: et al.: Genotypes of the cytochrome p450 47, 1331–1332 (2001). Determination of the mutation spectrum of isoform, CYP2C9, and the vitamin K 12. Gundry CN, Vandersteen JG, Reed GH, the EXT1/EXT2 genes in British Caucasian epoxide reductase complex subunit 1 Pryor RJ, Chen J, Wittwer CT: Amplicon patients with multiple osteochondromas, conjointly determine stable warfarin dose: melting analysis with labeled primers: and exclusion of six candidate genes in EXT aprospective study. J. Thromb. Thrombolysis a closed-tube method for differentiating negative cases. Hum. Mutat. 27, 1160 22, 191–197 (2006). homozygotes and heterozygotes. Clin. (2006). 3. Slinger R, Bellfoy D, Desjardins M, Chan F: Chem. 49, 396–406 (2003). 23. Zhou L, Wang L, Palais R, Pryor R, High-resolution melting assay for the •• First use of high-resolution Wittwer CT: High-resolution DNA melting detection of gyrA mutations causing melting analysis. analysis for simultaneous mutation scanning quinolone resistance in Salmonella enterica 13. Wittwer CT, Reed GH, Gundry CN, and genotyping in solution. Clin. Chem. 51, serovars Typhi and Paratyphi. Vandersteen JG, Pryor RJ: High-resolution 1770–1777 (2005). Diagn. Microbiol. Infect. Dis. 57, 455–458 genotyping by amplicon melting analysis using • Combined mutation scanning and (2007). LCGreen. Clin. Chem. 49, 853–860 (2003). genotyping by melting. 4. Wittwer CT, Kusukawa N: Real-time PCR. •• Introduction of saturating dyes. 24. Herrmann MG, Durtschi JD, Bromley LK, In: Diagnostic Molecular Microbiology; 14. Krypuy M, Newnham GM, Thomas DM, Wittwer CT, Voelkerding KV: Amplicon Principles and Applications. Persing DH, Conron M, Dobrovic A: High resolution DNA melting analysis for mutation Tenover FC, Versalovic J et al. (Eds). melting analysis for the rapid and sensitive scanning and genotyping: cross-platform ASM Press, Washington DC, USA 71–84 detection of mutations in clinical samples: comparison of instruments and dyes. Clin. (2004). KRAS codon 12 and 13 mutations in Chem. 52, 494–503 (2006). •• Technical review of real-time non-small cell lung cancer. BMC Cancer 6, • Instrument comparison for PCR methods. 295 (2006). melting analysis. 5. Wittwer CT, Herrmann MG, Moss AA, 15. Wojdacz TK, Dobrovic A: Methylation- 25. Herrmann MG, Durtschi JD, Bromley LK, Rasmussen RP: Continuous fluorescence sensitive high resolution melting (MS-HRM): Wittwer CT, Voelkerding KV: Instrument monitoring of rapid cycle DNA a new approach for sensitive and high- comparison for heterozygote scanning of amplification. Biotechniques 22, 130–131, throughput assessment of methylation. Nucleic single and double heterozygotes: 134–138 (1997). Acids Res. 35, e41 (2007). a correction and extension of Herrmann •• Introduction of SYBR Green I and 16. McKinney JT, Longo N, Hahn SH et al.: et al.: Clin. Chem. 2006; 52:494–503. Clin. hybridization probes to real-time PCR. Rapid, comprehensive screening of the Chem. 53, 150–152 (2007). 6. Wittwer CT, Ririe KM, Andrew RV, human medium chain acyl-CoA 26. Herrmann MG, Durtschi JD, Wittwer CT, David DA, Gundry RA, Balis UJ: The dehydrogenase gene. Mol. Genet. Metab. 82, Voelkerding KV: Expanded instrument LightCycler: a microvolume multisample 112–120 (2004). comparison of amplicon DNA melting fluorimeter with rapid temperature control. • First mutation scanning of a complex gene analysis for mutation scanning and Biotechniques 22, 176–181 (1997). with saturation dyes. genotyping. Clin. Chem. (2007) (In Press). 7. Ririe KM, Rasmussen RP, Wittwer CT: 17. Reed GH, Wittwer CT: Sensitivity and 27. Wittwer C, Kusukawa N: Molecular Product differentiation by analysis of DNA specificity of single-nucleotide diagnostics technology. In: Clinical melting curves during the polymerase chain polymorphism scanning by high-resolution Diagnostic Technology, The Total Testing reaction. Anal. Biochem. 245, 154–160 melting analysis. Clin. Chem. 50, Process, Volume 2: The Analytical Phase. (1997). 1748–1754 (2004). Ward-Cook K, Lehmann C, Schoeff L, • First use of fluorescent melting analysis for 18. Willmore C, Holden JA, Zhou L, Tripp S, Williams R (Eds). AACC Press, PCR product differentiation. Wittwer CT, Layfield LJ: Detection of Washington, DC, USA, 341–369 (2005). 8. Pirulli D, Boniotto M, Puzzer D, Spano A, c-kit-activating mutations in 28. Wittwer CT, Kusukawa N: Nucleic acid Amoroso A, Crovella S: Flexibility of gastrointestinal stromal tumors by techniques. In: Tietz Textbook of Clinical melting temperature assay for rapid high-resolution amplicon melting analysis. Chemistry and Molecular Diagnostics (4th detection of insertions, deletions, and Am. J. Clin. Pathol. 122, 206–216 (2004). Edition). Burtis C, Ashwood E, Bruns D single-point mutations of the AGXT gene 19. Zhou L, Vandersteen J, Wang L et al.: (Eds.). Elsevier, New York, NY, USA, responsible for type 1 primary High-resolution DNA melting curve 1407–1449 (2005). hyperoxaluria. Clin. Chem. 46, 1842–1844 analysis to establish HLA genotypic identity. • General overview of nucleic acid methods. (2000). Tissue Antigens 64, 156–164 (2004).

606 Pharmacogenomics (2007) 8(6) futurefuture sciencescience ggrouproup High-resolution DNA melting analysis for molecular diagnostics – REVIEW

29. Gingeras TR, Higuchi R, Kricka LJ, Lo YM, 41. Vandersteen JG, Bayrak-Toydemir P, 51. Dames SA, Margraf RL, Pattison D, Wittwer CT: Fifty years of molecular Palais RA, Wittwer CT: Identifying Wittwer CT, Voelkerding KV: Stabilty of 3´ (DNA/RNA) diagnostics. Clin. Chem. 51, common variants in heteroduplex scanning blocking modifications of unlabeled probes. 661–671 (2005). by high-resolution melting. Clin. Chem. 53. J. Mol. Diagn. (2007) (In Press). 30. Dujols VE, Kusukawa N, McKinney JT, (2007) (In Press). 52. Poulson M, Wittwer CT: Closed-tube Dobrowolski SF, Wittwer CT: 42. Margraf RL, Mao R, Highsmith WE, genotype of apo E with isolated probe PCR. High-resolution melting analysis for Holtegaard LM, Wittwer CT: Biotechniques (2007) (In Press). scanning and genotyping. In: Real-Time RET proto-oncogene genotyping using 53. Margraf RL, Mao R, Wittwer CT: Masking PCR. Dorak MT (Ed.). Garland Science, unlabeled probes, the masking technique, selected sequence variation by incorporating New York, NY, USA 157–171 (2006). and amplicon high-resolution melting mismatches into melting analysis probes. 31. Erali M, Palais B, Wittwer C: SNP analysis. J. Mol. Diagn. 9, 184–196 (2007). Hum. Mutat. 27, 269–278 (2006). genotyping by unlabeled probe melting 43. Chou LS, Lyon E, Wittwer CT: 54. Chou LS, Meadows C, Wittwer CT, analysis. In: Molecular Beacons – Signaling A comparison of high-resolution Lyon E: Unlabeled oligonucleotide probes Nucleic Acid Probes. Seitz O, Marx A (Eds). melting analysis with denaturing modified with locked nucleic acids for Humana Press, Totowa, NJ, USA. (2007) high-performance liquid chromatography improved mismatch discrimination in (In Press). for mutation scanning: cystic fibrosis genotyping by melting analysis. 32. Dobrowolski SF, Ellingson C, Coyne T transmembrane conductance regulator gene Biotechniques 39(5), 644 (2005). et al.: Mutations in the phenylalanine as a model. Am. J. Clin. Pathol. 124, 55. Lazaruk K, Walsh PS, Oaks F et al.: hydroxylase gene identified in 95 patients 330–338 (2005). Genotyping of forensic short tandem repeat with phenylketonuria using novel systems of 44. Seipp MT, Durtschi JD, Liew MA et al.: (STR) systems based on sizing precision in a mutation scanning and specific genotyping Unlabeled oligonucleotides as internal capillary electrophoresis instrument. based upon thermal melt profiles. Mol. temperature controls for genotyping by Electrophoresis 19, 86–93 (1998). Genet. Metab. (2007) (In Press). amplicon melting. J. Mol. Diagn. (2007) 56. Wang Y, Ju J, Carpenter BA, Atherton JM, 33. Lay MJ, Wittwer CT: Real-time (Epub ahead of print). Sensabaugh GF, Mathies RA: Rapid sizing fluorescence genotyping of factor V Leiden 45. Liew M, Nelson L, Margraf R et al.: of short tandem repeat alleles using capillary during rapid-cycle PCR. Clin. Chem. 43, Genotyping of human platelet antigens array electrophoresis and energy-transfer 2262–2267 (1997). 1 to 6 and 15 by high-resolution amplicon fluorescent primers. Anal. Chem. 67, • First use of probes for genotyping by melting and conventional hybridization 1197–1203 (1995). fluorescent melting analysis. probes. J. Mol. Diagn. 8, 97–104 (2006). 57. Schmalzing D, Koutny L, Chisholm D, 34. Bernard PS, Ajioka RS, Kushner JP, 46. Hill CE, Duncan A, Wirth D, Nolte FS: Adourian A, Matsudaira P, Ehrlich D: Wittwer CT: Homogeneous multiplex Detection and identification of cytochrome Two-color multiplexed analysis of eight genotyping of hemochromatosis mutations P-450 2C9 alleles *1, *2, and *3 by short tandem repeat loci with an with fluorescent hybridization probes. Am. high-resolution melting curve analysis of electrophoretic microdevice. Anal. Biochem. J. Pathol. 153, 1055–1061 (1998). PCR amplicons. Am. J. Clin. Pathol. 125, 270, 148–152 (1999). 35. Crockett AO, Wittwer CT: Fluorescein- 584–591 (2006). 58. Butler JM: Genetics and genomics of core labeled oligonucleotides for real-time PCR: 47. Odell ID, Cloud JL, Seipp M, Wittwer CT: short tandem repeat loci used in human using the inherent quenching of Rapid species identification within the identity testing. J. Forensic Sci. 51, 253–265 deoxyguanosine nucleotides. Anal. Biochem. Mycobacterium chelonae-abscessus group by (2006). 290, 89–97 (2001). high-resolution melting analysis of hsp65 59. Li L, Li CT, Li RY et al.: SNP genotyping by 36. Wittwer CT, Herrmann MG, Gundry CN, PCR products. Am. J. Clin. Pathol. 123, multiplex amplification and microarrays Elenitoba-Johnson KS: Real-time multiplex 96–101 (2005). assay for forensic application. Forensic Sci. PCR assays. Methods 25, 430–442 (2001). 48. Cheng JC, Huang CL, Lin CC et al.: Rapid Int. 162, 74–79 (2006). 37. Herrmann MG, Dobrowolski SF, detection and identification of clinically 60. Vaughn CP, Elenitoba-Johnson KS: Wittwer CT: Rapid β-globin genotyping by important bacteria by high-resolution High-resolution melting analysis for multiplexing probe melting temperature and melting analysis after broad-range ribosomal detection of internal tandem duplications. color. Clin. Chem. 46, 425–428 (2000). RNA real-time PCR. Clin. Chem. 52, J. Mol. Diagn. 6, 211–216 (2004). 38. Liew M, Pryor R, Palais R et al.: Genotyping 1997–2004 (2006). 61. Orita M, Iwahana H, Kanazawa H, of single-nucleotide polymorphisms by 49. Erali M, Pounder JI, Woods GL, Petti CA, Hayashi K, Sekiya T: Detection of high-resolution melting of small amplicons. Wittwer CT: Multiplex single-color PCR polymorphisms of human DNA by gel Clin. Chem. 50, 1156–1164 (2004). with amplicon melting analysis for electrophoresis as single-strand •• Introduction of small amplicon genotyping identification of Aspergillus species. Clin. conformation polymorphisms. Proc. Natl by high-resolution melting. Chem. 52, 1443–1445 (2006). Acad. Sci. USA 86, 2766–2770 (1989). 39. Palais RA, Liew MA, Wittwer CT: 50. Zhou L, Myers AN, Vandersteen JG, 62. Lerman LS, Silverstein K: Computational Quantitative heteroduplex analysis for single Wang L, Wittwer CT: Closed-tube simulation of DNA melting and its nucleotide polymorphism genotyping. Anal. genotyping with unlabeled oligonucleotide application to denaturing gradient gel. Biochem. 346, 167–175 (2005). probes and a saturating DNA dye. Clin. Electrophoresis Methods Enzymol. 155, 40. Graham R, Liew M, Meadows C, Lyon E, Chem. 50, 1328–1335 (2004). 482–501 (1987). Wittwer CT: Distinguishing different DNA •• Introduction of unlabeled probe 63. Xiao W, Oefner PJ: Denaturing heterozygotes by high-resolution melting. genotyping by melting and saturation dyes. high-performance liquid chromatography: Clin. Chem. 51, 1295–1298 (2005). areview. Hum. Mutat. 17, 439–474 (2001).

futurefuture sciencescience groupgroup www.futuremedicine.com 607 REVIEW – Reed, Kent & Wittwer

64. Li Q, Liu Z, Monroe H, Culiat CT: 68. Willmore-Payne C, Layfield LJ, Holden JA: melting amplicon analysis: correlation with Integrated platform for detection of DNA c-KIT mutation analysis for diagnosis of gene copy number, protein expression, and sequence variants using capillary array gastrointestinal stromal tumors in fine hormone receptor expression. Hum. Pathol. electrophoresis. Electrophoresis 23, needle aspiration specimens. Cancer 105, 37, 755–763 (2006). 1499–1511 (2002). 165–170 (2005). 72. Willmore-Payne C, Holden JA, Layfield LJ: 65. Bocker S: Simulating multiplexed SNP 69. Dobrowolski SF, McKinney JT, Detection of EGFR- and HER2-activating discovery rates using base-specific cleavage Amat di San Filippo C, Giak Sim K, mutations in squamous cell carcinoma and mass spectrometry. Bioinformatics 23, Wilcken B, Longo N: Validation of involving the head and neck. Mod. Pathol. e5–12 (2007). dye-binding/high-resolution thermal 19, 634–640 (2006). 66. Willmore-Payne C, Holden JA, Tripp S, denaturation for the identification of 73. Nomoto K, Tsuta K, Takano T et al.: Layfield LJ: Human malignant melanoma: mutations in the SLC22A5 gene. Hum. Detection of EGFR mutations in archived Detection of BRAF- and c-kit-activating Mutat. 25, 306–313 (2005). cytologic specimens of non-small cell lung mutations by high-resolution amplicon 70. Margraf RL, Mao R, Highsmith WE, cancer using high-resolution melting melting analysis. Hum. Pathol. 36, 486–493 Holtegaard LM, Wittwer CT: Mutation analysis. Am. J. Clin. Pathol. 126, 608–615 (2005). scanning of the RET protooncogene using (2006). 67. Willmore-Payne C, Holden JA, high-resolution melting analysis. Clin. 74. Willmore-Payne C, Volmar KE, Chadwick BE, Layfield LJ: Detection of Chem. 52, 138–141 (2006). Huening MA, Holden JA, Layfield LJ: c-kit exons 11- and 17-activating mutations 71. Willmore-Payne C, Holden JA, Layfield LJ: Molecular diagnostic testing as an adjunct to in testicular seminomas by high-resolution Detection of epidermal growth factor morphologic evaluation of pancreatic ductal melting amplicon analysis. Mod. Pathol. 19, receptor and human epidermal growth system brushings: Potential augmentation 1164–1169 (2006). factor receptor 2 activating mutations in for diagnostic sensitivity. Diagn. Cytopathol. lung adenocarcinoma by high-resolution 35, 218–224 (2007).

608 Pharmacogenomics (2007) 8(6) futurefuture sciencescience ggrouproup